COVID-19 RISK ASSESSMENT BASED ON SORE THROAT CONDITIONS BY USING MACHINE LEARNING

Abstract

The PCR test and antigen test are proven to provide a good way to screen COVID-19, however, those methods still required time which cannot be less than one minute and required high cost and special preparation. When the vaccine achieves 70% which means the population already in the herd immunity so, still this screening needs to be conducted because most properly that the COVID-19 will become endemic which never ends but still we need to conduct some assessment. There are some methods to conduct assessment either use saliva, sound and other methods but still, the problem in term of accuracy, cost and also speed, so in order to have this one of the methods that still challenging is to scan the sore throat and differentiate between the sore throat from normal flu and then COVID-19. In this research, we conducted a study to differentiate the image of a sore throat between a COVID-19 patient, flu patient, and normal patient. We collected the data from people having COVID-19, non-COVID-19 and normal people and then we conducted some pre-processing and then we apply two methods which are a conventional method and machine learning. As a test result, the system is able to differentiate the image from COVID-19, non-COVID-19, and normal people but the accuracy still having a problem because the data is limited but from the current result we saw there is potential to utilize the image from the sore throat to be used for the very restrict screening of the COVID-19. So, this system will be useful later on to be used in the airport, seaport, or others in order to reduce the risk of COVID-19 after people achieve herd immunity or 70% vaccinated.